Ever wondered how a college side project can turn into a $28 million AI marketing powerhouse? This week, we unpack how two UC Berkeley dropouts went from a cramped apartment to a venture-backed juggernaut— and the lessons you can apply to your own AI startup journey.

Why 46 percent of developers now distrust AI despite record adoption, how Reddit just overtook Wikipedia as the top-cited source in AI answers, and Meta’s $72 billion infrastructure play set to supercharge 2025.

From cramped apartments to $28M Series A through customer obsession and AI timing

The founding story of Conversion, an AI-powered marketing automation startup that raised $28M Series A led by Abstract Ventures. Founded by Neil Tewari (24) and James Jiao, two UC Berkeley dropouts who started with a $2M seed round at age 19 and are now approaching $10M ARR.

Learn how rigorous customer discovery (160 interviews with marketing VPs), frugal bootstrapping (seven roommates in a two-bedroom apartment), and perfect AI timing transformed a college side project into a venture-backed competitor to HubSpot and Marketo.

Build for yourself first, then validate obsessively — Conversion started as an internal tool to enhance HubSpot before the founders realized they'd solved a universal pain point in marketing automation workflows.

"We actually spent like two months doing like 160 customer interviews with VPs of marketing, 50- to 500-employee businesses, and got a much more positive response than we could have imagined."

Conversion targets businesses migrating from legacy tools (90% of customers "yanked out a legacy app") rather than first-time buyers— a strategic positioning that requires deeper sales cycles but higher customer value and stickiness.

The annual Stack Overflow survey of 49,000+ developers shows 84% now use AI tools (up from 76% in 2024), but trust has plummeted dramatically. Only 46% trust AI accuracy—a sharp increase from 31% distrust last year. Key findings reveal that 45% of developers find debugging AI-generated code time-consuming, and experienced developers show the lowest trust levels at just 2.6%.

Poe launched an API providing developers access to over 100 AI models across text, image, video, and voice. The API uses Poe's point-based subscription system, with developers able to purchase additional points at $30 per million tokens. This enables integration with tools like Cursor, Cline, and other OpenAI-compatible applications.

GitHub Spark, the AI-native app builder, entered public preview after being teased at GitHub's developer conference. The platform allows developers to build, iterate, and deploy web apps from natural language prompts, offering a more code-centric approach compared to AWS's similar Kiro platform.

YC-backed Decipher launched AI agents that watch user sessions in real-time to detect product issues and opportunities. The platform uses vision language models to identify problems from silent drop-offs to full errors, helping teams ship features instead of debugging session replays.

Reddit emerged as the most cited domain in AI-generated answers across all models, cited twice as often as Wikipedia in Q2 2025. Both Google AI Overviews and Perplexity rely heavily on Reddit content, while the platform now serves 70+ million weekly search users.

Meta revealed plans to spend up to $72 billion on AI infrastructure in 2025, escalating the compute arms race among tech giants. The investment focuses on physical and technical infrastructure needed to scale Meta's AI ambitions across its platforms.

DEV Community and Redis launched the Redis AI Challenge, calling on developers to build innovative AI-powered applications using Redis as a real-time data platform. Running through August 10, the hackathon encourages projects that explore Redis’ new AI and vector features, going beyond caching to power real-time search, LLM caching, and more. The challenge features $1,500 prizes, DEV++ membership, and community badges for valid submissions, making it a showcase of Redis’ evolving role in AI-centric application development.

Another week of separating AI signal from noise. If we saved you from a demo that would've crashed prod, we've done our job.

📧 Got a story? Reply with your AI tool wins, fails, or war crimes. Best stories get featured (with credit).

📤 Share the skepticism: Forward to an engineer who needs saved from the hype. They'll thank you.

✍️ Who's behind this? The Augment Code team—we build AI agents that ship real code. Started this newsletter because we're tired of the BS too.

🚀 Try Augment: Ready for AI that gets your whole codebase?